{"paper":{"title":"Towards reconstructing experimental sparse-view X-ray CT data with diffusion models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Diffusion priors trained on diverse synthetic data reconstruct experimental sparse-view X-ray CT scans better than narrow priors, with annealing reducing mismatch artifacts.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ezgi Demircan-Tureyen, Felix Lucka, Nelas J. Thomsen, Xinyuan Wang","submitted_at":"2026-02-13T09:33:39Z","abstract_excerpt":"Diffusion-based image generators are promising priors for ill-posed inverse problems like sparse-view X-ray Computed Tomography (CT). As most studies consider synthetic data, it is not clear whether training data mismatch (``domain shift'') or forward model mismatch complicate their successful application to experimental data. We measured CT data from a physical phantom resembling the synthetic Shepp-Logan phantom and trained diffusion priors on synthetic image data sets with different degrees of domain shift towards it. Then, we employed the priors in a Decomposed Diffusion Sampling scheme on"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our results reveal that domain shift plays a nuanced role: while severe mismatch causes model collapse and hallucinations, diverse priors outperform well-matched but narrow priors. Forward model mismatch pulls the image samples away from the prior manifold, which causes artifacts but can be mitigated with annealed likelihood schedules.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The physical phantom sufficiently resembles the synthetic Shepp-Logan phantom and that the Decomposed Diffusion Sampling scheme correctly balances the learned prior against the real forward model without introducing unaccounted biases in the experimental setting.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Diffusion priors trained on diverse synthetic data outperform narrow matched priors for experimental sparse-view CT reconstruction, but forward model mismatch introduces artifacts that annealed likelihood schedules can mitigate.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Diffusion priors trained on diverse synthetic data reconstruct experimental sparse-view X-ray CT scans better than narrow priors, with annealing reducing mismatch artifacts.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d3a910b2086e9dc88c113f5149945cc286e3f736587a18f9511c86949bb56d31"},"source":{"id":"2602.12755","kind":"arxiv","version":3},"verdict":{"id":"c1b51896-1b69-44e1-beab-7a153a9b7a1d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T22:39:42.136746Z","strongest_claim":"Our results reveal that domain shift plays a nuanced role: while severe mismatch causes model collapse and hallucinations, diverse priors outperform well-matched but narrow priors. Forward model mismatch pulls the image samples away from the prior manifold, which causes artifacts but can be mitigated with annealed likelihood schedules.","one_line_summary":"Diffusion priors trained on diverse synthetic data outperform narrow matched priors for experimental sparse-view CT reconstruction, but forward model mismatch introduces artifacts that annealed likelihood schedules can mitigate.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The physical phantom sufficiently resembles the synthetic Shepp-Logan phantom and that the Decomposed Diffusion Sampling scheme correctly balances the learned prior against the real forward model without introducing unaccounted biases in the experimental setting.","pith_extraction_headline":"Diffusion priors trained on diverse synthetic data reconstruct experimental sparse-view X-ray CT scans better than narrow priors, with annealing reducing mismatch artifacts."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.12755/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}